JP7067189B2 - Data processing method in glycohemoglobin analysis - Google Patents

Description

The present invention relates to a method for providing accurate quantitative calculation results in glycohemoglobin measurement by liquid chromatography.

Chromatography is used in various fields for the purpose of separating and quantifying a sample composed of various components. When performing chromatography for the purpose of quantification based on the area or height of each component peak, it is required to accurately calculate the peak area. Therefore, ideally, it is desirable to completely separate the individual components, but in reality, some components may not be completely separated.

Factors that disturb the shape of the peak include damage caused by the relationship between the target component and the column and the eluent, sample overload on the column, the effect of dead volume other than the direct separation field, and the temperature gradient in the column. Can be mentioned. In addition, when gradient elution is performed, the peak shape may not be symmetrical due to changes in the melt output over time. In particular, this phenomenon may become remarkable in "stepwise gradient elution" in which the eluent is gradually cut back. In addition to this, there are cases where the quantification accuracy is deteriorated due to the origin of the detector. In any detector, it is desirable that the signal output with respect to the concentration changes linearly, but the concentration range in which the linearity can be maintained is limited, and at higher concentrations, the rate of change in the output with respect to the concentration is large. It goes down and doesn't change. Accurate quantitative analysis cannot be performed at concentrations where the linearity is greatly disrupted.

Due to the various factors mentioned above, the peak shape of each component does not take a symmetrical shape even in the glycohemoglobin analysis. Further, since S-A1c% of a healthy person is 5 to 6% and A0 occupies about 90% of the whole, the intensity (peak height) of the A0 peak is extremely large, and S-A1c. In order to accurately calculate the area% of, it is necessary to accurately quantify the minute S-A1c peak and the huge A0 peak. That is, although the detector to be used is required to have a wide calibration range, there are individual differences in the patient sample, and the A0 peak may deviate from the calibration range, resulting in low reliability of the measurement result.

An object of the present invention is a process for obtaining an accurate peak area value even in a peak measured in a concentration region where separation is insufficient or the linearity of the detector is inferior in the measurement of glycohemoglobin by liquid chromatography. To provide a method.

The present invention has been made to solve the above-mentioned problems, and the peak shapes of other regions are mathematically calculated based on the region where the peak shape shows a theoretically correct shape, and the shape of the entire peak is formed. It is a method that accurately expresses and enables accurate quantitative calculation.

That is, the present invention has a non-normal distribution based on the output data of a part of the region constituting the peak of the chromatogram and the output data of the part of the region constituting the baseline of the chromatogram in the measurement of glycohemoglobin by liquid chromatography. Moreover, it is a chromatogram data processing method characterized in that curve fitting is performed using a function indicating a peak shape and quantitative calculation is performed using the result of the curve fitting.

As a function having a non-normal distribution and a peak shape (hereinafter, may be referred to as a "processing function"), known functions such as the following (Equation 1) to (Equation 5) can be used, but these functions Not limited to. Note that y is an output, x is a time, and xc, y0, A, w, w1, w2, and w3 are coefficients, respectively.

First, the procedure of the present invention will be described.
(Step 1)
Select a processing function and specify the initial value of the corresponding coefficient. As initial values, xc is the time of the apex of the peak to be processed, y0 is the base output value of the peak to be processed, A is the output value (peak height) of the peak to be processed, and w, w1, w2, w3 are. It is good to use the half width of the peak to be processed as a guide.
(Step 2)
Two or more fitting intervals are specified for the peaks (including the peak group) of the chromatogram to be processed. The order of the step 1 and the main step may be reversed, and the order is not particularly limited.
(Step 3)
In the fitting section, the difference from the calculated value of the processing function for which the initial value is specified in the step 1 and then the square of the difference are calculated, and the sum of the squares of the differences (hereinafter referred to as "error") is obtained.
(Step 4)
The coefficient of the processing function is sequentially changed so that the error is minimized, and the optimum coefficient is obtained. This step can be easily performed by using the "solver" installed in the spreadsheet software Excel2010 of Microsoft Corporation or software having a similar function.
(Step 5)
The peak group is fitted using the coefficient of the processing function with the minimum error, and the quantitative (area) calculation is performed.

Next, a case where the present invention is applied to the A0 peak will be described.

As shown in FIG. 1b, the output of the detector with respect to the concentration may slow down from a certain concentration to the change of concentration. When the linearity is good, a peak shape close to the normal distribution can be obtained, but when the linearity deteriorates from a certain concentration, the peak shape becomes dull near the peak apex as shown in FIG. 1a, and the A0 peak is exactly that. It corresponds to such a peak.

Assuming that the peak output (detector output) is 75 or more and the linearity is lost as shown in FIG. 2, the time (horizontal axis) is 0 to 1.4 minutes and 2.55 to 5.0 minutes. A linear relationship is obtained between the concentration and the output, and there is no linear relationship between the concentration and the output for 1.4 to 2.55 minutes. In such a case, although the peak shape should be as shown by the broken line in FIG. 2a, the signal actually obtained is as shown by the solid line. The peak area based on the shape of the solid line is calculated to be smaller than the peak area based on the shape of the broken line. Therefore, the shape of the entire peak is predicted by calculation using the region (data point cloud of the rising part of the peak and the falling part of the peak) where the actually obtained peak shape and the ideal peak shape are the same. , Carry out area calculation (quantitative calculation).

In the case of Fig. 2, the peak rising part is from 0 minutes to 1.4 minutes at the minimum, but for the purpose of eliminating the influence of baseline drift etc., the linearity of the output starts from 0.72 minutes when the peak actually rises. It is specified to be up to 1.4 minutes near the upper limit that can be maintained (shaded area). Similarly, the peak falling part is from 2.55 minutes to 5.0 minutes at the maximum, but it is specified from 2.55 minutes near the upper limit where linearity can be maintained to 3.2 minutes when the peak actually falls (the peak falls). Diagonal area). Further, in order to specify the absolute position of the baseline, a plurality of points in the region before the peak and the region after the peak where the baseline does not fluctuate are defined (FIG. 2b). Fitting is performed by the above-mentioned procedure with these data groups as the fitting section. Table 1 shows the results when the coefficients are sequentially changed by using the asymmetric double sigmoid function as the processing function.

FIG. 2c is a diagram schematically showing an approximate curve with the optimum parameter (try3) obtained based on the above operation (broken line in the figure). As can be seen from FIG. 3, there is almost no deviation in the fitting section, and it can be seen that the optimum fitting is performed. In the vicinity of the peak peak (1.5 to 2.0 minutes), the dissociation between the calculated curve and the original chromatogram is conversely large, and this also shows the ideal peak shape in this region. You can also see that it wasn't.

Next, a case where the present invention is applied to the SA1c peak will be described.

The SA1c peak is a group of peaks as shown in FIG. 4, which is composed of a principal component and a plurality of other components and is insufficiently separated.

As shown in FIG. 4, a large peak that becomes the main component (first component) at 2.0 minutes, a small peak that becomes a minute component (second component) at 1.0 minute, and a minute component (third component) at 3.2 minutes. Assuming that there is a small peak that becomes a component), the first component and the second component are from 0.5 minutes to 1.3 minutes, and the first component is from 2.5 minutes to 3.9 minutes. The component and the third component are mixed. On the other hand, the vicinity of the apex of the peak is a region of a single component of the first component which is the main component. In this case, it is estimated that about 1.5 to 2.45 minutes is only the first component.

Therefore, as shown in FIG. 5, the shape of the entire peak is predicted by calculation using the region where the actually obtained peak shape and the ideal peak shape are the same, that is, the data point cloud near the peak peak. Then, perform area calculation (quantitative calculation). The calculation area is defined from 1.5 minutes, which is estimated to be free of the minute second component, to 2.45 minutes, which is estimated to be free of the minute third component (shaded area). Further, in order to specify the absolute position of the baseline, a plurality of points in the region before the peak and the region after the peak where the baseline does not fluctuate are defined (FIG. 5b). Fitting is performed by the above-mentioned procedure with these data groups as the fitting section. Table 2 shows the results when the coefficients are sequentially changed by using the asymmetric double sigmoid function as the processing function.

FIG. 5c is a diagram schematically showing an approximate curve with the optimum parameter (try3) obtained based on the above operation (broken line in the figure). As can be seen from FIG. 6, there is almost no dissociation in the calculation interval, and it can be seen that the optimum fitting is performed. In the vicinity of the base of the peak (0.8 to 1.3 minutes, 2.7 to 3.7 minutes), the dissociation between the curve calculated in reverse and the original chromatogram is large, which is also the reason for this. It can be seen that the region suggests that minute components other than the main component are mixed.

In the glycohemoglobin measurement by liquid chromatography, accurate peak area values can be obtained even for peaks with insufficient separation and peaks measured in a concentration region where the linearity of the detector is poor.

It is a figure which showed typically the single peak in the state where the detector output is saturated. It is a figure explaining the case which carries out the calculation process of this invention with respect to a single peak. It is a figure which showed the final result (error minimum point) when the calculation process of this invention is performed on a single peak. It is a figure which showed typically the peak shape which consists of a plurality of components, and the constituent peak group. It is a figure explaining the case where the calculation process of this invention is performed on the peak group consisting of a plurality of components. It is a figure which showed the final result (error minimum point) when the calculation process is performed on the peak group consisting of a plurality of components. It is a figure which showed the system configuration which performed Examples 1 and 2. It is a figure which showed the separation pattern of standard glycohemoglobin and the report output used for comparison in Examples 1 and 2. It is a chromatogram of the sample used in Example 1. It is a figure which showed the chromatogram at the dilution ratio 1/40 in Example 1. FIG. It is a figure which plotted the area value of A0 peak with respect to the dilution rate in Example 1. FIG. It is a figure which showed the composition of each component in the glycohemoglobin separation chromatogram in Example 2 simply. 6 is a chromatogram showing an approximation result of the SA1c peak in Example 2. It is a figure which showed the correlation of SA1c area% in Example 2 and SA1c area% obtained by the general peak detection (vertical cutting) under the standard condition which separation is good.

Hereinafter, the present invention will be specifically shown with reference to Examples, but the present invention is not limited to this Example.

An asymmetric double sigmoid function was used as the processing function used for fitting, and the calculation was performed using the spreadsheet software Excel2010 manufactured by Microsoft as a means for minimizing the error by iterative calculation. "Solver", which is one of the functions of this spreadsheet software, is a function that can obtain the optimum value of a variable in order to obtain a target value in a mathematical formula including a plurality of variables. We used the solver function to search for parameters to minimize the error.

First, the initial value of the coefficient of the function was input. For xc, a rough peak elution time was specified, and a constraint was set to change it within a range of ± 10%. The time and output of the chromatogram to be fitted were input, the solution at the corresponding x (time) was obtained, and the square of the difference between the actual output and the calculated solution was calculated. The sum of the squares (errors) of the differences in each interval was calculated, and the solver function was used to change the coefficients so that the errors were minimized, and repeated calculations were performed.

(Example 1)
Glycohemoglobin was measured by liquid chromatography and applied to the A0 peak.

As a measuring device, Tosoh automatic glycohemoglobin analyzer HLC-723GVIII was used. FIG. 7 shows the main system configuration.

In the case of this apparatus, the blood sample is separated by a column by a three-component step gradient method, and the separated hemoglobin component is detected by the absorbance at a wavelength of 415 nm. The three-component step gradient is a method in which three types of eluents are switched to stepwise using an ion exchange column to separate each component, and eluents 1, 0. From sample injection to 0.11 minutes (T1). Eluent 2 is sent from 11 minutes (T1) to 0.53 minutes (T2), and eluent 3 is sent from 0.53 minutes (T2) to 0.63 minutes (T3), and the eluate 3 is returned to eluent 1.

FIG. 8a is a typical chromatogram (report) obtained with the Tosoh automatic glycohemoglobin analyzer HLC-723GVIII. FIG. 8b is an enlarged chromatogram near the SA1c peak, and FIG. 8c is a chromatogram showing the whole. As can be seen from these figures, there is a difference of about 50 times in peak height between the S-A1c peak and the A0 peak, which are the main purposes.

In the SA1c measurement, a blood sample (whole blood) is usually diluted with a hemolysis / diluted solution to about 1/150 for measurement. In this example, the blood sample (whole blood) is diluted 1/40, 1/50, 1/60, 1/80, 1/100, 1/150, 1/200, 1/250. Diluted and measured. The obtained chromatogram is shown in FIG. The figure on the left is an enlarged view of the vicinity of the S—A1c peak, and the figure on the right is a diagram showing the entire area so that the A0 peak shape and the output value can be understood by fixing the output axis.

Comparing the highest concentration dilution ratio 1/40 (FIG. 9a) with the general concentration dilution ratio 1/150 (FIG. 9b), it can be seen that the peak shape of A0 is significantly different. The A0 peak with a dilution rate of 1/150, which is a general concentration, is sharp, while the A0 peak with a dilution rate of 1/40, which has the highest concentration, is distorted near the apex.

Linearity can be guaranteed up to about 1500 mV of the detector signal, and it is assumed that the detector signal approaches saturation above that, and the data point group of 1500 mV or less at the rising part of the A0 peak and 1500 mV or less at the falling part of the A0 peak. In order to specify the absolute position of the data point group and baseline of the data point group, the four regions of multiple points, the region before the peak and the region after the peak, where there is no fluctuation in the baseline, are set as fitting sections, and the dilution rate with high concentration is 1/40. The present invention was applied to a chromatogram having a dilution rate of 1/50 and a dilution rate of 1/60. The coefficients of the optimized processing function are as follows.

FIG. 10a is a chromatogram of a sample having a dilution ratio of 1/40 and data points to be calculated, and FIG. 10b is an enlarged view of the vicinity of A0. FIG. 10c is a diagram showing the A0 peak shape obtained by the above treatment. The solid line shows the untreated chromatogram, and the broken line shows the shape of A0 obtained by the calculation result. As can be seen from the figure, the output (height) of the A0 peak obtained by this calculation process is about 500 mV larger than that of the unprocessed A0 peak, and the distortion of the peak shape is also reduced.

FIG. 11 is a diagram in which the area based on the A0 peak shape obtained above and the area of the A0 peak in the untreated chromatogram are plotted against the dilution ratio. As can be seen from this, the area of A0 obtained from the raw data (chromatogram) cannot be guaranteed to be linear when the dilution ratio of the sample is 1/60 or more, but it is linear by performing the treatment of the present invention. Sexual improvement was seen.

As a result, SA1c% based on the area of A0 treated with the present invention can be calculated more accurately even in such a high concentration range.

(Example 2)
Glycohemoglobin was measured by liquid chromatography in the same manner as in Example 1 and applied to the SA1c peak.

In this embodiment, in order to verify whether SA1c% can be calculated accurately even in a state where the separation of the L-A1c to SA1c peaks and the SA1c to A0 peaks is insufficient, the above-mentioned eluent switching is performed. A chromatogram in which the stepwise time and the flow rate were adjusted and the measurement time was shortened by 40% was performed.

FIG. 12 is an example of the chromatogram used in this verification. Under this condition, the SA1c peak elutes at about 0.42 minutes and the A0 peak elutes at about 0.50 minutes. As can be seen in comparison with the chromatogram under the standard conditions of FIG. 9, it can be seen that the separation of the L-A1c to S-S1c peaks and the S-S1c to A0 peaks is poor.

From 0.410 to 0.453 minutes (data points: 25 points), including the SA1c peak top, and to identify the absolute position of the baseline, the pre-region and peak of the peak with no baseline variation. The present invention was applied by designating a data point cloud in which a plurality of points in the rear region were calculated as a fitting section (see FIG. 13).

In order to verify the accuracy of this treatment, comparison with the value obtained under the condition of clinically measuring A1c% was carried out. As a comparison target, SA1c% of Tosoh Corporation glycohemoglobin analyzer GHbVIII in the standard mode was used (analysis time: 1.0 minute). FIG. 14 shows the correlation between the chromatogram with insufficient separation (analysis time of about 0.6 minutes) used in this example and SA1c% calculated by applying the data processing method of the present invention. The horizontal axis plots SA1c% obtained in the GHbVIII standard mode (separation is good and the peak vertical cutting process is performed), and the vertical axis plots the SA1c% obtained in this example. In addition, 7 whole blood samples, controls (high and low values), and standard (high and low values) were used (each n = 2 measurements).

Since the correlation formula is y = 0.9923x-0.003, R ^ 2 is 0.9991, the slope of the correlation formula is almost 1, the intercept is also extremely small, and R ^ 2 is also very close to 1. Both values indicate that the correlation is very good and that almost the same values are obtained.

From the above, it was confirmed that even if the present invention is applied to a chromatogram of glycohemoglobin whose separation is not good, S-A1c% can be calculated accurately.

1. 1. Eluent A
2. 2. Eluent B
3. 3. Eluent C
4. Hemolysis / diluted solution 5. Deaerator 6. Solenoid valve 7. Liquid feed pump 8. Specimen dilution / injection device 9. Preheating coil 10. Pre-filter 11. Analytical column 12. Oven 13. Visible detector 14. Waste liquid

Claims (4)

In the glycohemoglobin measurement by liquid chromatography, the non-normal distribution is based on the output data of the rising and falling parts of the A0 peak that constitutes the peak of the chromatogram and the output data of a part of the region that constitutes the baseline of the chromatogram. A chromatogram data processing method comprising performing curve fitting using a function indicating a peak shape and performing quantitative calculation using the result of the curve fitting. In glycohemoglobin measurement by liquid chromatography, the peak is non-normally distributed based on the output data of the maximum part of the SA1c peak that constitutes the peak of the chromatogram and the output data of a part of the region that constitutes the baseline of the chromatogram. A chromatogram data processing method comprising performing curve fitting using a function indicating a shape and performing quantitative calculation using the result of the curve fitting. In glycohemoglobin measurement by liquid chromatography, the output data of the rising and falling parts of the A0 peak that constitutes the peak of the chromatogram, the maximum part of the SA1c peak, and a part of the region that constitutes the baseline of the chromatogram. A chromatogram data processing method characterized in that curve fitting is performed using a function showing a non-normal distribution and a peak shape based on output data, and quantitative calculation is performed using the result of the curve fitting. The data processing method according to any one of claims 1 to 3, wherein the function having a non-normal distribution and showing a peak shape is an asymmetric double sigmoid function.

Images